AI-UX Process and Practice

Last updated June 24, 2024

To understand how to design from an AI-UX first perspective, I joined a startup called Verdi AI preparing an MVP of a Product Manager copilot. I worked with the CEO, developers, and prompt engineers to imagine probabilistic workflows that complemented junior product managers to develop, de-risk, and validate their ideas into stronger business cases.

Though that project flopped for funding reasons, my concept of complementary AI and human-AI relationships deepened as I researched, synthesizing my findings in my blog Systems At Play, focused on systems thinking and burning design questions.

Views on AI are differing depending on a person’s perspective or motives, so I’m creating to zoom out and see the big picture of where design can sculpted a beneficial, complementary future with AI. I also vibe coding and doing AI-UX projects to experience each tool’s limits.

Tools I use:

  • ChatGPT

  • Claude

  • NotebookLM

  • Stitch

  • Cursor

  • Replit

Writing what I learn

  • Perspective: Seeing AI as a Good Ancestor

    Read

  • Solving System Problems with Limited Time, Resources, and Understanding

    Read

  • Overview of Agentic AI

    Coming soon

  • Designing the Human-AI Relationship

    Coming soon
Probabilistic decision-making can be available but it isn't useful unless the user is choosing the right problem to solve. (Opportunity Graph feature for Verdi)

Probabilistic decision-making can be available but it isn't useful unless the user is choosing the right problem to solve. (Opportunity Graph feature for Verdi)

How should we guide the Human-AI relationship?

  • Coming soon

  • Coming soon

  • Coming soon

Creating a relationship that enhances quality of life focuses on improving interfaces that enable more time, energy, joy, love, and peace.

Takeaways so far

My 3-month stint at Verdi (Dec 2023-2024) taught me that AI is not a strategy but a tool that must be grounded in a clear problem and strategy.

Vibe coding has limitations in incorporating design consistency and specific user interactions, but it can help people build finer knowledge of technical capabilities.

Reasoning models are excellent starting points for deep research. Input all your questions and the POV you want it to view from. Read a comprehensive report to get started. Having a quick understanding of a domain makes it easier to have context.

Next

Enabling eMortgage Closing and Investing from Homebuyer to Big Banks